DK2635048T3 - The gain of a voice signal in response to the input level - Google Patents

Description

Description

The present invention relates to a method for amplifying a speech signal for a hearing apparatus by detection of an input sound by the hearing apparatus, determining an input level of the input sound and amplification of the input sound by the hearing apparatus. The present invention also relates to a hearing apparatus having a detection device for detecting an input sound, a measuring device for determining an input level of the input sound and an amplification device for amplifying the input sound. A hearing apparatus is here taken to mean any device which can be worn in or on the ear and which generates an acoustic stimulus, in particular a hearing aid, headset, headphones and the like.

Hearing aids are wearable hearing apparatuses that are used to support the hard of hearing. Different hearing aid designs, such as behind-the-ear hearing aids (BTE), hearing aids with an external receiver (RIC: receiver in the canal) and in-the-ear hearing aids (ITE), for example also concha hearing aids or completely-in-canal hearing aids (ITE, CIC) are provided in order to accommodate the numerous individual requirements. The hearing aids listed by way of example are worn on the outer ear or in the auditory canal. However, bone conduction hearing aids, implantable or vibrotactile hearing aids are also commercially available, moreover. In this case damaged hearing is either mechanically or electrically stimulated.

In principle hearing aids have as their fundamental components an input converter, an amplifier and an output converter. The input converter is usually a sound pick-up, for example a microphone and/or an electromagnetic receiver, for example an induction coil. The output converter is usually implemented as an electroacoustic converter, for example a miniature loudspeaker, or as an electromechanical converter, for example a bone conduction receiver. The amplifier is conventionally integrated in a signal processing unit. This basic construction is shown in Fig. 1 using the example of a behind-the-ear hearing aid. One or more microphones 2 for receiving the sound from the environment are fitted in a hearing aid case 1 for wearing behind the ear. A signal processing unit 3, which is also integrated in the hearing aid case 1, processes the microphone signals and amplifies them. The output signal of the signal processing unit 3 is transmitted to a loudspeaker or receiver 4 which outputs an acoustic signal. The sound is optionally transmitted via a sound tube, which is fixed to an otoplastic in the auditory canal, to the eardrum of the wearer of the aid. The energy supply to the hearing aid, and in particular that of the signal processing unit 3, takes place by way of a battery 5 likewise integrated in the hearing aid case 1.

The dynamic range of hearing is significantly reduced in the case of a central hearing impairment. One solution to this problem lies in using hearing aids with compression. Compression in hearing aids has the function of transferring the dynamic range of people with normal hearing into that of the people with hearing difficulties. What is involved in particular here is the shifting of quiet signals into the audible range, but of not amplifying loud signals any further. Dynamic variations are a problem in this connection.

Retention or improvement of the intelligibility when wearing hearing aids is usually the primary aim. A great challenge lies therefore in finding a suitable reduction in the dynamic range for the type of speech signal in particular. The problem in this connection lies in particular in the variations over time in speech levels. Both slow changes in speech level, which reproduce the variations in the overall speech level, and fast changes in speech level, which reproduce the level differences between speech components (vowels, consonants, phonemes), are important.

Signal processing of hearing aids is known from the unpublished German application DE 10 2005 061 000 A1 in which an input signal is classified with respect to the current hearing situation. The input signal is amplified in accordance with a first compression algorithm or a second compression algorithm as a function of the classification result.

The aim of some known compression strategies with slow time constants is to maintain the rapid fluctuations in level over time between vowels and consonants. The quiet parts of speech (in particular between vowels) are often inaudible as a result or simply too quiet with this type of compression.

The object of the present invention therefore lies in also making quiet parts of speech audible without significantly raising the overall level of the speech signal.

According to the invention this object is achieved by a method for amplifying a speech signal for a hearing apparatus by detection of an input sound by the hearing apparatus, determining an input level of the input sound and amplification of the input sound by the hearing apparatus, wherein a mean broadband input level of the input sound is determined, it is established whether the input sound has a speech component, and if not, the input signal is amplified as a function of the input level in accordance with a predefined first amplification characteristic, if so, the input signal is amplified as a function of the input level in accordance with a predefined second amplification characteristic, wherein the second amplification characteristic is formed as a function of the mean broadband input level such that amplification is increased with respect to the first amplification characteristic in a predefined level interval below the mean broadband input level. An amplification characteristic is the function of amplification over the input level here. “Broadband” is here taken to mean the overall spectral bandwidth of signal processing of the hearing apparatus.

According to the invention a hearing apparatus is also provided, having a detection device for detecting an input sound, a measuring device for determining an input level of the input sound and an amplification device for amplifying the input sound, wherein a mean broadband input level of the input sound can be determined by the measuring device, it can be established by a classification device of the hearing apparatus whether the input sound has a speech component, and the amplification device can be controlled by the classification device in such a way that the input signal is amplified as a function of the input level in accordance with a predefined first amplification characteristic if no speech component is found in the input signal, and the input signal is amplified as a function of the input level in accordance with a predefined second amplification characteristic if a speech component is found in the input signal, wherein the second amplification characteristic is formed as a function of the mean broadband input level such that amplification is increased with respect to the first amplification characteristic in a predefined level interval below the mean broadband input level.

An adaptive elevation of speech signals advantageously occurs therefore as a function of input level. This is achieved in that the amplification in a predetermined range below the mean broadband current input level, which is roughly the level of vowels, is higher in the case of a detected speech signal than in the case where no speech signal is detected. The level of consonants in particular may be increased as a result, so, overall, intelligibility increases significantly.

The predefined level interval is preferably a mean interval between a vowel level and a consonant level at the input. The vowel level is the level for which the mean can be measured in the case of vowels in a current speech signal. The consonant level is the level for which the mean is measured in the case of consonants in current speech. Since the vowel level substantially corresponds to the overall speech level and is therewith easy to measure, the consonant level can be easily determined with the known level interval from the measured vowel level. The mean interval between vowel level and consonant level is known for average speakers in the case of loud and quiet speech.

The second amplification characteristic can lie in a level range of predefined breadth above the first amplification characteristic. This means that the second amplification characteristic lies not only above the first amplification characteristic in the case of a fixed input level, but in a greater range of input levels. This range of input levels should cover the fundamental level range of the consonants. The predefined breadth preferably lies roughly in the order of magnitude of the predefined level interval. If the predefined breadth roughly corresponds to the level interval between consonants and vowels and the elevation takes place roughly symmetrically to the mean consonant level, the range of elevation roughly extends to the centre between consonant level and vowel level.

The values of the second amplification characteristic advantageously lie at a maximum of 12 dB to 24 dB above those of the first amplification characteristic. Intelligibility can best be increased with this elevation between a maximum of 12 and 24 dB.

The method is advantageously applied solely to sound components of the input sound with frequencies above a predefined frequency, e.g. 1500 Hz, in particular above 3000 Hz. This is because consonants have significant frequency components above 1500 Hz. With respect to intelligibility it is substantially sufficient if frequency components are elevated above 3000 Hz using the inventive method.

The second amplification characteristic can be increased differently with respect to the first amplification characteristic in various frequency bands. The speech signal components are elevated as a function of frequency in addition to as a function of input level. This increases intelligibility and the natural sound impression is consequently better retained as well.

It is also advantageous if amplification is performed in a first path and a second path, parallel thereto, of the hearing apparatus, a first compression is performed in the first path and a second compression is performed in the second path, the second compression is greater than the first compression, and in the second path amplification takes place in accordance with the first and second amplification characteristic. An advantageous, non-linear compression can be achieved by the overlaying of two different compression characteristics. In particular the signals of the first and second paths can be mixed with a mixing ratio which depends on the input level of the input sound. The compression can be dynamically adjusted to the current input sound thereby. At the same time, intelligibility may be increased when using the different amplification characteristics for speech signals and non-speech signals.

The present invention will be illustrated in more detail with the aid of the accompanying drawings, in which:

Fig. 1 shows the schematic construction of a hearing aid according to the prior art;

Fig. 2 shows a graph of an amplification characteristic over the input level in the case of first speech;

Fig. 3 shows an amplification characteristic over the input level in the case of second speech;

Fig. 4 shows a parallel compression as a function of a mean input level in a hearing aid and;

Fig. 5 shows an electrical circuit diagram for the implementation of a parallel compression.

The exemplary embodiments described in more detail below are preferred embodiments of the present invention.

It has been found that intelligibility depends inter alia on the ratio of the consonant level to the vowel level in the speech signal of an individual speaker. In this connection the consonant level means the mean level of consonants and the vowel level the mean level of vowels in speech. Intelligibility can be improved by increasing the consonant level, i.e. the level of the consonants, without simultaneously increasing the vowel level (reference level). The improvement in intelligibility achieved in this way is then at its highest if the consonant-vowel ratio (CVR; i.e. interval between the consonant level and the vowel level) is higher than in the case of average speech. Particularly good results with respect to in telligibility have been produced if the consonant level is elevated or increased with respect to the vowel level by up to 12 to 24 dB. This has advantages in particular in the case of consonants in the higher frequency spectrum, i.e. above 1500 Hz and in particular above 3000 Hz. Primarily affected by this are fricatives (“s”, “sh”, “f”, etc.) and what are known as stop consonants (“p”, “t”, etc.).

To be able to benefit from this knowledge, a special compression strategy must be applied which is performed in addition to “normal” compression. A strategy of this kind can be illustrated with the aid of Fig. 2 for a first speech A. For this purpose the hearing apparatus or hearing aid must have a classifier or similar device which can establish whether the input sound has a speech component or not. If a speech component is not detected in the input sound, a level-dependent amplification occurs in accordance with a first amplification characteristic 10 (shown in Fig. 2 as a solid line). It can be seen that the amplification g of the input level L, depends on the input sound. If, on the other hand, a speech component is found in the input sound, level-dependent amplification occurs in accordance with a second amplification characteristic 11. This is partially on the first amplification characteristic 10 and partially above it.

The second amplification characteristic is always adjusted to current speech A (cf. Fig. 2) or B (cf. Fig. 3). For this purpose the overall speech level of current speech A or B is measured broadband in a first step (overall processing bandwidth). This overall speech level, which is also called the mean broadband input level here, is dominated by the vowel level V. This means that the overall speech level primarily depends on the volume of the vowels. The ratio of the volumes of consonants to vowels (CVR) is also known. This ratio CVR corresponds to the interval between the vowel level V and the consonant level C. The ratio CVR can be determined by way of a plurality of measurements and averaging. The consonant level C can therefore easily be derived from the overall signal level, which roughly corresponds to the vowel level, by subtracting a standard or mean CVR from the measured overall speech level. A current, significantly lower consonant level C then results for current speech according to the measured overall speech level or vowel level. In the example of Fig. 2 the vowel level V is measured therefore for speech A, and the consonant level C lies in an interval CVR below it. In the example of Fig. 3 the current overall speech level or vowel level V’ is measured for speech B, so consonant level C’ results minus the standard CVR.

In the case of the determined consonant level C or C’ amplification g is now increased with respect to the first amplification characteristic 10. However, amplification is increased not only exactly at value C or C’ of input level U but also in a region R or R’ in each case around the respective value C or C’. The respective second amplification characteristic 11 or 1T then results with this elevation, and this lies, at least in certain regions (in the regions R and R’), above the first amplification characteristic 10.

The first amplification characteristic 10 and the second amplification characteristic 11, 11’ preferably run steadily. In other words, they do not exhibit any jumps. The amplification characteristics can be composed of linear sections in this case, as in the examples of Fig. 2 and Fig. 3. The amplification characteristics can also have non-linear courses, however. The respective courses in the amplification characteristics may also be achieved by standard expansion methods and standard compression methods.

In the examples of Fig. 2 and 3 a maximum elevation m or m’ is performed at input level C or C’, i.e. at the mathematically determined consonant level. This maximum elevation should lie between 12 and 24 dB. The best results in relation to intelligibility may be achieved with these elevation values. The maximum elevations need not lie at C or C’ exactly, however. The respective elevation region in relation to C or C’ need not be symmetrical either. It is essential only that C or C’ lies in the elevation region R or R’.

The compression elevation should only be applied in the case of high-frequency speech components above 1500 Hz and in particular above 3000 Hz. This affects what are known as high-frequency consonants therefore, whose funda mental frequency components lie above said frequencies. These are fricatives and stop consonants in particular.

As will be illustrated with the aid of Fig. 4 and 5, the elevation of the amplification in the case of the consonant levels C and C’ may be linked to what is known as parallel compression. The problem of simple compression, which operates either very slowly or very quickly, may be remedied by this parallel compression. With very slowly readjusted amplification it is not possible in particular to amplify very quiet details without providing too much amplification for louder signals. If, by contrast, adjustment works very quickly, quiet details become audible. Overall the sound suffers a great deal, however, since the dynamics of natural signals are completely lost. With parallel compression on the other hand, the signal is split onto two channels 12, 13. One channel 13 compresses the signal very moderately and very slowly in accordance with a compression characteristic 14 to primarily avoid output variables which too loud. The other, parallel, channel 12 compresses the input signal very strongly and very quickly in accordance with a compression characteristic 15. At the end the signals of both paths 12, 13 are added again in an adder 16. The mixing ratio of the two channels 12, 13 is determined as a function of a mean input level L averaged over time. The aim of adaptive mixing is to compress as little as possible in the case of very quiet environments, but relatively highly in the case of loud levels. The situation is therefore avoided where quiet background noises from the environment are amplified too much in the silence. Overall, a resulting compression characteristic 17 is then produced which, by way of example at low input levels, is dominated by fast compression 15 and at higher input levels by slow compression 14.

The aim of this compression method is to retain the sound quality of the input signal as far as possible despite high and fast compression. The clarity of a signal depends to a large extent on pulses which structure the sound pattern. Conventional, fast compression rounds these pulses. The parallel compression method can be parameterised in such a way that the louder pluses are always added to the output signal uncompressed, whereas quiet details can be added via the other channel. The compressed pulses contained therein are covered by those in the uncompressed or very slow channel. Precisely this parameterisa-tion can be easily achieved due to the dependency of the mixing ratio on the mean input level.

If speech is then detected in the input signal, the compression characteristic 15 for fast compression can be replaced by a compression characteristic which is based on the second amplification characteristic 11 or 1T (cf. Fig. 2 and 3). An additional elevation would result thereby in region C of the input level U Such an additional elevation 18 is symbolised in Fig. 4 by a dot-dash line. This elevation 18 would be part of the resulting compression characteristic 17 in the case of speech components.

The exemplary embodiments mentioned have the advantage that by using the known interval CVR, specifically the amplification of consonants in the case of compression can easily be increased, so ultimately intelligibility may be improved. In addition, there is an adaption to the current overall level of speech with compression. The interval CVR can be specified in this connection, e.g. for quiet, medium and loud speech.

There is also no abrupt increase in the consonant level, but a gentle transition instead. An abrupt transition in amplification would have the adverse effect of excessively amplifying ambient background noise. The gentle transition, by contrast, smoothes the sound quality and renders the method more robust against other background noises.

Claims (8)

1. Fremgangsmåde til forstærkning af et talesignal til et høreapparat ved hjælp af - registrering afen indgangslyd ved hjælp af høreapparatet, - bestemmelse af indgangslydens indgangsniveau (Li), og - forstærkning af indgangslyden ved hjælp af høreapparatet, hvorved - et gennemsnitlig bredbåndet indgangsniveau (V, V’) for indgangslyden bestemmes, - konstatering af, om indgangslyden omfatter en taledel, og - hvis nej, foregår forstærkningen af indgangssignalet i afhængighed af indgangsniveauet ifølge en på forhånd bestemt første forstærkningskarakteristik (10), - hvis ja, foregår forstærkningen af indgangssignalet i afhængighed af indgangsniveauet (Li) ifølge en på forhånd givet anden forstærkningskarakteristik (11, 11’), kendetegnet ved, at den anden forstærkningskarakteristik er således tildannet i afhængighed af det gennemsnitlige bredbåndede indgangsniveau (V), at i en på forhånd given niveauafstand (CVR) under det gennemsnitlige bredbåndede indgangsniveau (V) er forstærkningen (g) hævet i forhold til den første forstærkningskarakteristik (10), hvorved den på forhånd givne niveauafstand (CVR) er en gennemsnitlig afstand imellem et vokalniveau og et konsonantniveau (C), og hvorved den anden forstærkningskarakteristik (11, 11’) ligger i et niveauområde med på forhånd given bredde (R, R’) over den første forstærkningskarakteristik (10).1. A method of amplifying a speech signal to a hearing aid by means of: - recording an input sound by means of the hearing aid, - determining the input sound level (Li), and - amplifying the input sound by means of the hearing aid, whereby - an average broadband input level (V (V ') for the input sound is determined, - ascertaining whether the input sound comprises a speech part, and - if no, the amplification of the input signal is dependent on the input level according to a predetermined first gain characteristic (10), - if yes, the amplification of the input signal takes place. in dependence on the input level (Li) according to a predetermined second gain characteristic (11, 11 '), characterized in that the second gain characteristic is so formed in dependence on the average broadband input level (V) that at a predetermined level distance ( CVR) below the average broadband input level (V) is the gain (g) raised relative to the first gain characteristic (10), whereby the predetermined level distance (CVR) is an average distance between a vocal level and a consonant level (C), and thereby the second gain characteristic (11). , 11 ') lies in a level region of predetermined width (R, R') above the first gain characteristic (10). 2. Fremgangsmåde ifølge krav 1, hvorved den på forhånd givne bredde (R, R’) ligger i størrelsesordenen af den på forhånd givne niveauafstand (CVR).The method of claim 1, wherein the predetermined width (R, R ') is in the order of the predefined level distance (CVR). 3. Fremgangsmåde ifølge et af de foregående krav, hvorved værdien af den anden forstærkningskarakteristik (11, 11’) ligger maksimalt 12 dB til 24 dB over den første forstærkningskarakteristiks (10) værdi.Method according to one of the preceding claims, wherein the value of the second gain characteristic (11, 11 ') is a maximum of 12 dB to 24 dB above the value of the first gain characteristic (10). 4. Fremgangsmåde ifølge et af de foregående krav, hvorved den anvendes udelukkende til lyddele af indgangslyden med frekvenser over en på forhånd given frekvens.Method according to one of the preceding claims, wherein it is used exclusively for sound portions of the input sound at frequencies above a predetermined frequency. 5. Fremgangsmåde ifølge et af de foregående krav, hvorved den anden forstærkningskarakteristik (11, 11 ’) er hævet forskelligt i forskellige frekvensbånd i forhold til den første forstærkningskarakteristik (10).Method according to one of the preceding claims, wherein the second gain characteristic (11, 11 ') is raised differently in different frequency bands relative to the first gain characteristic (10). 6. Fremgangsmåde ifølge et af de foregående krav, hvorved forstærkningen udføres i en første (12) og en hermed parallel anden sti (13) i høreapparatet, der i den første sti (12) gennemføres en første kompression, og i den anden sti en anden kompression, den anden kompression er stærkere end den første kompression, og forstærkningen i den anden sti (13) sker i overensstemmelse med den første eller anden forstærkningskarakteristik (10, 11, 11’).A method according to any one of the preceding claims, wherein the reinforcement is performed in a first (12) and a parallel second path (13) in the hearing aid, a first compression is carried out in the first path (12) and in the second path a second compression, the second compression is stronger than the first compression, and the gain in the second path (13) occurs according to the first or second gain characteristic (10, 11, 11 '). 7. Fremgangsmåde ifølge krav 6, hvorved den første og anden stis (12, 13) signaler blandes med et blandingsforhold, som afhænger af indgangslydens indgangsniveau (Li).The method of claim 6, wherein the first and second path (12, 13) signals are mixed with a mixing ratio which depends on the input level (L1) of the input sound. 8. Fløreapparat med - en registreringsindretning til registreringen afen indgangslyd, - en måleindretning til bestemmelsen af et indgangsniveau (Li) for indgangslyden, og - en forstærkningsindretning til forstærkning af indgangslyden, hvorved - der ved hjælp af måleindretningen er konstaterbar en gennemsnitlig båndbredde af indgangsniveauet (V) for indgangslyden, - der af en klassifikationsindretning for høreapparatet kan konstateres, om indgangslyden omfatter en taledel, kendetegnet ved, at - forstærkningsindretningen styres således at klassifikationsindretningen, at ° indgangssignalets forstærkning foregår i afhængighed af indgangsniveauet (Li) ifølge en på forhånd given første forstærkningskarakteristik (10), hvis der ikke konstateres nogen taledel i indgangssignalet, og ° forstærkningen af indgangssignalet foregår i afhængighed af indgangsniveauet (Li) ifølge en på forhånd givet andet forstærkningskarakteristik (11, 11’), hvis der konstateres en taledel i indgangssignalet, hvorved den anden forstærkningskarakteristik er således tildannet i afhængighed af det gennemsnitlige bredbåndede indgangsniveau (V), at i en på forhånd given niveauafstand (CVR) under det gennemsnitlige bredbåndede indgangsniveau (V) hæves forstærkningen (g) i forhold til den første forstærkningskarakteristik (10), hvorved den på forhånd givne niveauafstand (CVR) er en gennemsnitlig afstand imellem et vokalniveau og et konsonantniveau (C), og hvorved den anden forstærkningskarakteristik (11, 11’) ligger over den første forstærkningskarakteristik (10) i et niveauområde af på forhånd given bredde (R, R’).8. An apparatus with - a recording device for recording an input sound, - a measuring device for determining an input level (Li) for the input sound, and - an amplifier for amplifying the input sound, whereby - by means of the measuring device an average bandwidth of the input level is ascertained ( V) for the input sound, - it can be ascertained by a classifier for the hearing aid whether the input sound comprises a speech part, characterized in that - the amplifier is controlled so that the classifier means that ° the amplification of the input signal is dependent on the input level (Li) according to a predetermined first level gain characteristic (10) if no speech portion is detected in the input signal, and ° the gain of the input signal is dependent on the input level (Li) according to a predetermined second gain characteristic (11, 11 ') if a speech portion is found in the input signal, whereby the second gain characteristic is formed in dependence on the average broadband input level (V) that, at a predetermined level distance (CVR) below the average broadband input level (V), the gain (g) is raised relative to the first gain characteristic. (10), wherein the predetermined level spacing (CVR) is an average distance between a vowel level and a consonant level (C), and wherein the second gain characteristic (11, 11 ') is above the first gain characteristic (10) in a level range of preset width (R, R ').